Bourbaki

Exercise 1, § 5

Tag 03MD, permanent. Cite it as [Bourbaki, Tag 03MD].

Theory of Sets, III, 235

Prove the formula

∑k=q+1n−p+q+1(n−kp−q−1)(k−1q)=(np),\sum_{k=q+1}^{n-p+q+1} \binom{n-k}{p-q-1}\binom{k-1}{q} = \binom{n}{p},

where p⩽np \leqslant n and q<pq < p (generalize the argument of no. 8, Corollary to Proposition 14).

Solution

This solution was written by nemotron-3-ultra-free and judged by a machine. It is not Bourbaki's and it has not been checked by a person.

Let nn, pp, qq be integers such that p⩽np \leqslant n and q<pq < p. Consider the set E=[1,n]E = [1, n] of integers from 11 to nn. By Corollary 1 to Proposition 11, the number of subsets of EE with exactly pp elements is (np)\binom{n}{p}.

For each integer kk with q+1⩽k⩽n−p+q+1q+1 \leqslant k \leqslant n-p+q+1, let Pk\mathscr{P}_k be the set of all pp-element subsets XX of EE whose (q+1)(q+1)-th smallest element equals kk. The sets Pk\mathscr{P}_k are pairwise disjoint and their union is the set of all pp-element subsets of EE; indeed, if XX is a pp-element subset with elements x1<⋯<xpx_1 < \dots < x_p, then xq+1x_{q+1} must satisfy q+1⩽xq+1⩽n−(p−q−1)=n−p+q+1q+1 \leqslant x_{q+1} \leqslant n - (p-q-1) = n-p+q+1, so XX belongs to exactly one Pk\mathscr{P}_k.

Fix kk in the above range. By Proposition 5 of no. 3, the interval [1,k−1][1, k-1] has k−1k-1 elements and the interval [k+1,n][k+1, n] has n−kn-k elements. To form a subset X∈PkX \in \mathscr{P}_k, we must choose qq elements from [1,k−1][1, k-1] and p−q−1p-q-1 elements from [k+1,n][k+1, n], and then take their union with {k}\{k\}. The number of ways to choose qq elements from [1,k−1][1, k-1] is (k−1q)\binom{k-1}{q} by Corollary 1 to Proposition 11 applied to the set [1,k−1][1, k-1]. The number of ways to choose p−q−1p-q-1 elements from [k+1,n][k+1, n] is (n−kp−q−1)\binom{n-k}{p-q-1} by the same corollary applied to [k+1,n][k+1, n]. These choices are independent, and the mapping (A,B)↦A∪B∪{k}(A, B) \mapsto A \cup B \cup \{k\} is a bijection from the Cartesian product of the set of qq-element subsets of [1,k−1][1, k-1] and the set of (p−q−1)(p-q-1)-element subsets of [k+1,n][k+1, n] onto Pk\mathscr{P}_k. Hence ∣Pk∣=(k−1q)(n−kp−q−1)|\mathscr{P}_k| = \binom{k-1}{q} \binom{n-k}{p-q-1}.

Since the sets Pk\mathscr{P}_k form a partition of the set of all pp-element subsets of EE, the cardinal of the latter is the sum of the cardinals of the Pk\mathscr{P}_k. Therefore

(np)=∑k=q+1n−p+q+1(k−1q)(n−kp−q−1), \binom{n}{p} = \sum_{k=q+1}^{n-p+q+1} \binom{k-1}{q} \binom{n-k}{p-q-1},
which is the required formula.

Cited by

Nothing.

Cites

Transcribed from the printed volume. Not translated.